Page 109 - Code Craft Computer-8

3. What value will the output variable contain?

These types of questions are asked when the output (target) variable is continuous, which means it
may take any value within a range. For example, you are trying to predict the stock prices of a
particular company on the basis of its historical data. The price can be any value, such as ₹20,
₹20.5, ₹21, ₹25, ₹27, and so on. It means the output is not constrained to any specific value. Some
more examples of this kind of situation are:
• What will the temperature be like today and tomorrow?

• What would the price of a house in a specific area be?
• How much rain will fall tomorrow?

• What would the stock price be?
• What will be the sales of product for next month?

Classification algorithms do not work well for these types of problems. Regression algorithms in
machine learning are used to answer these kinds of questions.
Regression is a method that is based on mathematical formulas and helps in finding relationships
between different features of a dataset.

4. Which technique should be used to group the data?

In some applications, you are not aware of output variables. You only know the features; in this
case, data may be separated by creating different groups on the basis of some parameters or
features. For example , consider a box where there are multiple objects of three shapes (triangle,
rectangle, and circle). If you ask a three-year-old to segregate these objects, he will make three
groups (one for each shape), but he/she may not be aware of the names of those groups or shapes.
It means that he/she is aware that different shapes have different features, but their names are not
known to him/her.

For this kind of data, the clustering method is used.
In clustering, groups are formed on the basis of features similarities.

The type of machine learning used in the above scenario is called unsupervised machine learning.
In this technique, the output or the target variable is not known.

5. How to decide the next step?
Normally, this question arises in the case of autonomous vehicles, robots, and drones. Because
their decisions depend on the external environment, every time they need to work with the

external world. Reinforcement machine learning techniques are used to solve these kinds of
problems.
Reinforcement learning is also a type of machine learning in which learning is based on rewards
and punishment.

109